A novel approach for differentiation of liquid samples with surface acoustic wave transducers and embedded microcavities

Sensors and Actuators B 196 (2014) 272–281

Contents lists available at ScienceDirect

Sensors and Actuators B: Chemical journal homepage: www.elsevier.com/locate/snb

A novel approach for differentiation of liquid samples with surface acoustic wave transducers and embedded microcavities Sukru U. Senveli a,c,∗ , Onur Tigli a,b,c a b c

Electrical and Computer Engineering, University of Miami, Coral Gables, FL 33146, USA Department of Pathology, Miller School of Medicine, University of Miami, Miami, FL 33136, USA Dr. John T. Macdonald Foundation, Biomedical Nanotechnology Institute at University of Miami, Miami, FL 33136, USA

a r t i c l e

i n f o

Article history: Received 29 October 2013 Received in revised form 31 January 2014 Accepted 3 February 2014 Available online 12 February 2014 Keywords: Surface acoustic waves (SAW) Rayleigh waves Acoustic and fluid interaction

a b s t r a c t We discuss a novel method for sensing and differentiation of analytes trapped in a microcavity with an emphasis on liquids. The proposed sensing mechanism relies on capturing the analyte of interest in a microcavity etched on the delay line in contrast to the conventional mass loading method. The structure mainly consists of input and output interdigitated transducer (IDT) electrodes in an otherwise standard delay line configuration operated in Rayleigh mode along with a microcavity etched between the IDTs to trap minute amounts of liquids. Firstly, the responses of the system with the microcavity are explored using finite element method (FEM) analysis. Then, experimental results from delay lines on two different substrates, namely, Y-Z lithium niobate and ST-X quartz are analyzed. The system can distinguish between liquids with glycerin concentrations ranging from 60% to 90% in water and less than 5 pL in volume in the high frequency range of 197 MHz and 213 MHz based on frequency and phase shift readings. Lithium niobate samples with 1.2 ␮m deep microcavities provide an overall frequency sensitivity of −7.7 kHz/(% glycerin). Quartz samples with 8.5 ␮m deep microcavities have a sensitivity of −0.13◦ /(% glycerin). The minimum density–viscosity product experimentally differentiated using embedded micro√ cavities is 1.9 kg/m2 s. It is concluded that this method can be used to trap and interrogate minute amounts of liquids with different properties. Experimental results demonstrate that our approach can possibly be extended to certain solids, and to more complex structures like single biological cells. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Surface acoustic waves (SAW) are a form of acoustic waves that travel along the surface of a medium. Their properties are altered to great extents upon interaction with irregularities along their path of propagation; therefore, they make excellent candidates for sensors. Traditionally, acoustic wave sensors have been studied extensively and used in fields as diverse as temperature, torque, and viscoelasticity measurements in various modes such as pressure, mass loading, and viscosity [1,2]. SAW gas sensors and biosensors are generally used in mass loading mode where the added acting mass of the delay line medium or any particulates captured in the delay line portion of the system causes a drop in the resonance frequency of the filter configuration [3–5]. Surface functionalization coatings and absorbent films

∗ Corresponding author at: Electrical and Computer Engineering, University of Miami, Coral Gables, FL 33146, USA. Tel.: +1 305 284 6055. E-mail addresses: [email protected], [email protected] (S.U. Senveli). http://dx.doi.org/10.1016/j.snb.2014.02.007 0925-4005/© 2014 Elsevier B.V. All rights reserved.

usually aid the capturing process in the case of biosensors. Rayleigh waves are one of the most commonly used types of surface waves. The fact that their energy is contained close to the surface makes the device surface very sensitive at the expense of the capability of immersed operation due to extensive loading losses from liquids. On the other hand, Lamb wave, Love and other SH-wave based devices have been used for liquid based measurements and specific measurements such as viscosity [6–8]. As a result, research trends moved in the direction where Rayleigh mode SAW sensors did not find much use for integration with liquids. Still, some attempts have been made in the past where large masses of liquids were used with bounce and waveguide modes for coupling with Rayleigh waves [9]. There have been plenty of studies toward acoustic streaming of liquids with or without particles and acoustic processors [10]. Two things are common in most of these studies. Firstly, amounts of liquid being used are quite large, causing complete conversion of SAWs to acoustic compressional waves in liquid thereby eliminating the response due to SAWs on the receiving end. Secondly, such wave interactions have generally been considered as manipulation or actuation tools as opposed to sensing tools with some exceptions [11].

S.U. Senveli, O. Tigli / Sensors and Actuators B 196 (2014) 272–281

The approach we outline in this study differs from the previous studies outlined above in both categories, and uses SAW–liquid interactions as a sensing and differentiation mechanism for minute amounts of material. In the past, the only similar research on a similar topic with embedded microcavities has been a theoretical study about insertion of rectangular prism shaped polystyrene plugs into the delay line of SH-wave sensors with the purpose of not sensing but improving transmission characteristics and loading sensitivity [12]. No other research exists on SAW interactions for interrogation of minute amounts of matter in a similar way to our approach to the best of our knowledge. Here we present a novel sensing mechanism in which a material, in this case, a very small amount of liquid, placed inside a microcavity etched on the delay line can be interrogated using Rayleigh type surface waves. The mechanism differs from standard delay line configurations as an embedded microcavity which breaks the delay line is introduced. This structure causes dispersion and loss of a portion of SAW to bulk waves depending on the geometry but the resultant output transfer characteristics are dependent on the way the surface wave interacts with the material trapped in the microcavity. It is aimed to show that the properties of minute amounts of samples can be interrogated using this approach. During the course of this paper, first we give the overview of the model we use for the system followed by finite element method (FEM) analysis. After detailing the microfabrication procedures, experimental techniques are discussed. Finally, measurement results are presented.

2. Modeling and simulations The study of piezoelectric–liquid interface has been carried out in the literature before, particularly in the discussion of acoustic streaming applications. In the commonly studied versions of this setup, the interface between the piezoelectric layer and liquid droplet is only on the lateral plane where a liquid droplet remains on the surface of the piezoelectric layer with a contact angle given by the hydrophobicity/hydrophilicity of the substrate. After hitting the boundary, the Rayleigh wave decays rapidly and the transverse component emits longitudinal waves into the liquid at a certain angle. This critical angle is defined in an analogous fashion to Snell’s law, where it is defined in terms of the ratio of the wave velocities in the two media [13]. On the other hand, detailed solutions and exact analytical expressions for Rayleigh waves can be found in [14]. In our case, a similar approach is used, however, there is a microcavity filled with a liquid in the delay line and the surface wave penetrates into it according to a pressure based acoustics model. This acoustics model treats the pressure propagating inside the liquid according to a wave equation. An overview of the domains involved is shown in Fig. 1(a). A simplified 2D version of the actual problem to be solved is depicted in Fig. 1(b). Simulations were carried out using commercial finite element method (FEM) simulation package, Comsol Multiphysics. The simulation efforts can be broken down into three sections. Firstly a regular delay line is modeled to form a base model for the SAW IDTs after the correct surface wave mode has been identified. The defining characteristic of SAW behavior, insertion loss information, is obtained. Following this, response with an empty microcavity placed in the delay line is obtained. Then, simulation results where the microcavity is filled with different liquids are presented. In the simulations, the liquid has been modeled as a Newtonian fluid defined by its density, characteristic speed of sound, and viscosity. The simulation domains cover the contact of liquid and the piezoelectric surfaces which are connected to each other via appropriate boundary conditions. Also, it is assumed that there is no slip associated with the liquid interface. The interface between piezoelectric–air domains and liquid–air domains can be considered

273

Fig. 1. (a) A schematic representation of the important domains in the model. The characteristics of the media and independent variables are indicated. (b) Depiction of the 2D version of the real problem to be solved with the cavity. The air–liquid and air–piezoelectric boundaries can be simplified and neglected due to negligible acoustic transmission coefficient between these two pairs of media, hence eliminating the air domain completely.

as a soft boundary for acoustic waves where the pressure is zero as the acoustic transmission from relatively high acoustic impedance media to very low impedance medium such as air can be safely neglected. The Rayleigh wave mode is found around 215 MHz for Y-Z lithium niobate and 204 MHz for ST-X quartz using harmonic analysis to serve as a basis for further frequency domain simulations. Wave velocities are calculated as 3446 m/s and 3259 m/s, respectively which are close to the expected and tabulated phase velocity of these waves on the Y-Z lithium niobate, 3488 m/s, and on ST-X quartz, 3158 m/s. The 2-D simulation of the system can provide insight into the behavior and coupling of the piezoelectric and liquid domain, as long as the microcavity is large enough or the aperture is small enough (in the out-of-plane direction) so that they overlap. If the microcavity and aperture are close in size, the effects of non-uniformity along the out-of-plane axis which gives rise to diffraction artifacts can be ignored. In simulating the system, we use constant overlap electrode pairs and unity metallization ratio to simplify the designs. The output insertion loss characteristics of the system can be obtained in one of two ways: directly from frequency domain simulations or by applying Fourier transform to the time domain results to convert them to frequency domain data (which, in practice, is applying discrete Fourier transform to a good approximation). The frequency domain solutions were preferred over time domain solutions in general as they are less time consuming and they allow for direct examination and evaluation of the exact wave behavior at a given frequency as shown in Fig. 2. The IDTs in the simulations contain 64 electrode pairs, whereas the electrode width and spacing are both 4 ␮m yielding a wavelength of 16 ␮m. This wavelength is selected since it is comparable to the microcavity diameter. The number of electrode pairs yields a sharp enough passband peak to comfortably observe the output while keeping the simulated system size manageable. On the other hand, the microcavity diameter is selected to be 24 ␮m or 1.5 as it is easy to work with and also this size allows for further testing of similar sized objects in the micro scale for future studies. It is also an easier task to fabricate actual SAW microcavity sensors with corresponding feature sizes. A design trade off in the proposed devices is the size of the aperture. As the aperture is made smaller,

274

S.U. Senveli, O. Tigli / Sensors and Actuators B 196 (2014) 272–281

Fig. 2. Snapshot from the simulated system showing the response of the system excited at its peak frequency with the launched SAW traveling in +x direction. The color map corresponds to displacement for the piezoelectric substrate and pressure distribution for the liquid domain as shown with the scale bars. (For interpretation of the references to color in text, the reader is referred to the web version of this article.)

the interaction is constrained specifically to the microcavity which gives rise to increased transmission losses for SAW and increases errors especially on substrates with comparatively lower coupling constants such as quartz. If the aperture is larger, the transmission loss characteristics are improved but a smaller fraction of the aperture is covered by the microcavity, meaning that any interaction in the microcavity has a smaller effect on the output. Note that the microcavities are not made larger on purpose as the motive behind this study is to interrogate micro-sized material samples. For this reason, only apertures very close to the microcavity size should be considered. Simulation of the system with empty microcavities is imperative so that the effect of the fluid can be understood. In these simulations with microcavities, surface waves still propagate on the microcavity boundaries, i.e. the sidewalls and bottom surface of the cavity. Inclusion of this discontinuity is seen not to deteriorate the output characteristics to a significant amount unless the microcavity geometry causes extraneous conversion to bulk waves as shown in Fig. 2. Simulations show that small microcavity depths compared to the wavelength do not have a significant effect on the regular device operation as shown in Fig. 3. The liquid filling in the microcavity causes a resonant condition within the microcavity which is dependent on the geometry of the microcavity. The next section discusses the difficulties such as volatility and solidification in dispensing liquids in miniscule droplets. These difficulties limited us to work with glycerin and deionized (DI) water mixtures. The effect of the microcavity filling on the sensor performance was analyzed using these liquid mixtures in simulations. No slip condition was implicitly assumed for liquid–piezoelectric interfaces. Parametric simulations were run mainly for the given density, sound speed, and viscosity of liquids as given in Table 1. The parameters were adapted from [15,16] for various concentrations glycerin/water mixtures and variations of these parameters are given in Fig. 4. For increasing glycerin concentration, the increase in the viscosity is exponential whereas density and sound velocity have closer to linear relationships. In the experiments, two different substrates are considered for practical purposes as lithium niobate is very difficult to etch

beyond a small depth. For this reason, the simulation studies with liquids are concentrated on shallow microcavities (1 ␮m) with lithium niobate whereas deeper microcavities are used with quartz (8 ␮m). Using two cavity depths allows us to evaluate the difference between the cases where losses to the bulk are prominent and where they are not. The simulations were carried out in frequency domain, although the formation and time evolution of these waves have been affirmed with time domain simulations. There are two sources to consider regarding the resultant wave and the output. Firstly, the surface wave, propagating along the microcavity walls and the bottom surface, emerges from the other side of the cavity. Secondly, the transverse component of this SAW interacts with the contents of the microcavity to emit acoustic radiation and thus form acoustic waves inside the microcavity, in a similar fashion to acoustic streaming approaches. These acoustic waves are confined in a resonant cavity type configuration within the microcavity and couple back to the piezoelectric domain. This is the second source of SAW on the other side of the cavity. The combination of the first and second waves forms a resultant wave which is detected by the output IDT. Ultimately, the output signal is expected to be related to the properties of the liquid in the microcavity as shown in Fig. 5. Simulation data from a droplet on a regular delay line on ST-X quartz is provided for comparison to the system with an embedded microcavity. The trend observed in both cases with microcavities is decreasing phases due to the delays the liquids impose on the traveling SAW. In the shallow microcavity case, frequency shifts point to an operation mode more similar to mass-loading. While a similar trend is observed in the deep microcavity case, it is not as linear as its shallow counterpart. Insertion loss magnitudes in either case are not to be considered as a primary figure of merit as the losses to bulk are governing the transmission from that perspective. Observation of the case in Fig. 5(c) with a standard delay line reveals a smaller sensitivity compared to the microcavity case in Fig. 5(b). Microcavities increase the observable interaction with liquid which results in a higher phase sensitivity. So in essence, incorporation of the microcavity on the delay line increases the length of the boundary with the liquid causing a larger delay. This

Table 1 Properties of liquid samples used in the simulations and experiments. Sample number

L1

L2

L3

L4

Glycerin volume ratio in glycerin: DI water mixture (%) Glycerin mass ratio in glycerin: DI water mixture (%) Density (kg/m3 ) Sound velocity (m/s) Viscosity (mPa s) √ Density–viscosity product (kg/m2 s)

60 65.5 1171 1763 14.0 4.0

70 74.7 1195 1795 30.0 6.0

80 83.5 1218 1817 75.7 9.6

90 91.9 1241 1869 241.7 17.3

S.U. Senveli, O. Tigli / Sensors and Actuators B 196 (2014) 272–281

275

Fig. 3. Comparison of insertion loss characteristics with 24 ␮m diameter empty microcavities of varying depths in SAW devices with 64 electrode pairs on (a) Y-Z lithium niobate substrate and (b) ST-X quartz substrate.

Fig. 5. Simulation results from (a) Y-Z lithium niobate with 1 ␮m deep microcavities, (b) ST-X quartz with 8 ␮m deep microcavities, and (c) ST-X quartz standard delay line with no microcavities. The volume of the liquid in (c) is the same as the microcavity filled with liquid in (b) in order to provide a fair comparison. Liquids 1–4 are 60%, 70%, 80%, and 90% glycerin, respectively. Fig. 4. Properties of glycerin/water mixtures at room temperature. The liquids (L1–L4) selected for testing are shown.

276

S.U. Senveli, O. Tigli / Sensors and Actuators B 196 (2014) 272–281

is in addition to another important benefit, that is, helping trap a certain amount of liquid which would otherwise be difficult to precisely measure using most injection and dispensation systems available today. Although incorporating microcavities is beneficial for these reasons, too deep microcavities, for instance, those deeper than 0.75 generally cause large drops in insertion loss and result in strong reflections along with significant mode conversion to bulk waves. This decreases the interaction of the Rayleigh surface wave with the liquid and is detrimental to system performance. The complexity of the system makes it difficult to define a figure of merit for evaluating the performance as there are quite a few correlated variables at play. It was seen from the simulations that increasing the density of the liquid has a very similar effect in comparison to standard mass loading mode operation as expected if the microcavity depth is small. However, with deeper microcavities, more complicated resonant microcavity conditions occur. The most reliably correlated parameter that arises from the simulations is the phase angle of the output. The phase angle provides a measure of how much delay is introduced by the liquid component of the system. In order to define a figure of merit for the system, the simple Newton–Laplace formulation of acoustic impedance can be considered. This relation defines the acoustic

Fig. 6. Simulation results quartz with 8 ␮m deep microcavities with varying diameters. Liquids 1–4 are 60%, 70%, 80%, and 90% glycerin, respectively.

Fig. 7. Fabrication flow for microcavity based SAW sensors. SEM images of the third step of both layers of fabrication are shown (the metal layers are intact in both SEM images).

S.U. Senveli, O. Tigli / Sensors and Actuators B 196 (2014) 272–281



impedance of a liquid at a certain frequency as Zaco = jω where √ j is −1, ω is the angular frequency,  is the density, and  is viscosity. Therefore the acoustic impedance is directly proportional √ to , the density–viscosity product. The results obtained from simulations show that the acoustic impedance of the liquid can be indirectly measured using the microcavity approach due to the resonance condition. As the liquids with larger acoustic impedances closer to those of the substrates being used, smaller shifts are observed. However it should be noted that the system is cyclic in nature. That is to say, after a certain impedance value depending on that of the substrate is passed, the system will complete the phase angle reversal due to the cyclic nature of the phase (only defined between −180◦ and 180◦ ) and start again. Therefore, the system has a limited dynamic range due to its resonance condition. It was observed in the simulations that deeper microcavities incur more limited ranges. The highest sensitivity, i.e. the largest slope for the output phase curve, was found for a microcavity with /2 depth among other depths for the substrates considered through simulations; however, controllable etching of lithium niobate has not been attainable in a fashion in our experiments or in the literature. For this reason, lithium niobate has only been considered for simulations with very shallow cavities filled with liquids. Due to simplicity of its etching, quartz is simulated with deeper microcavities which are attainable in experiments. The microcavity can be designed to be wider for higher sensitivity but this also increases the attenuation of the Rayleigh wave and breaks the resonance condition inside the cavity. Simulations show that microcavities larger than 1.5 begin to show higher phase shifts but the relation between liquid concentration and phase difference is lost as too many wavelengths now fit inside the microcavity. If the microcavity is kept small with respect to wavelength the correlation between liquid concentration and phase difference holds but the sensitivity is lower as a much smaller portion of the wavelength fits in the microcavity. In order to keep the volume small and still comparable to other possible analytes of interest in the future, obtain a negative phase slope with increasing

277

concentration for all the four liquids used, and to obtain reasonable sensitivity, the width was selected as 1.5 and the depth as /2 (Fig. 6). 3. Materials and methods Y-Z cut lithium niobate (LiNbO3 ) and ST-X cut quartz (SiO2 ) samples were obtained from Boston Piezooptics. The former substrate is preferred because of its high coupling constant. However, etching deep microcavities in this substrate while maintaining low surface roughness to support SAW propagation is difficult. The quartz substrate is preferred due to easier fabrication and temperature stability although its coupling constant is much lower in comparison. The fabrication flow is shown in Fig. 7. Prior to the process steps, the samples were first cleaned in acetone for 20 min using a 20 kHz ultrasonic buzzer. After a short dehydration period at 110 ◦ C in an oven with nitrogen flow, samples were sputtered with a 1.0 ␮m layer of aluminum to serve as a hard mask for substrate etching step. AZ5214E photoresist (PR) was used in positive mode in lithography and the samples were then etched with a commercially available aluminum etchant. Microcavities can be opened by either wet or dry etching for both substrates although difficulties arise for lithium niobate substrates in either case. Etching lithium niobate is a very difficult process which is outside the scope of this study. For this reason, only shallow microcavities were produced on lithium niobate. Among the RIE recipes, the most aggressive one was found to be SF6 based etching by applying 200 W RF power and a pressure of 70 mT. Due to excessive heat build-up, the sample chuck was constantly cooled and etching was carried out intermittently to avoid overheating. Under these conditions the etch rate was found to be approximately 0.2 ␮m/h. The results of this etch is shown in Fig. 7. Etching of quartz was carried out also with RIE but with a CF4 /O2 recipe at 300 W RF power under 170 mT of pressure. No visible damage was observed on the substrate surface due to aluminum etching for opening microcavities or due to removing of the hard mask layer. The depths of the microcavities on lithium

Fig. 8. Substrates after fabrication is completed along with SEM images of some large aperture SAW devices.

278

S.U. Senveli, O. Tigli / Sensors and Actuators B 196 (2014) 272–281

Fig. 9. Dispensing liquids into the microcavity using probes, sequentially from (a) to (e).

niobate and quartz were measured as 1.2 ␮m and 8.5 ␮m, respectively, whereas the microcavity diameter is 24 ␮m for both substrates. For patterning the electrodes and IDT structures, lift-off process was preferred. The alignment step was carried out with specifically designed alignment marks to ensure correct orientation of the

crystal for SAW propagation. AZ5214E type photoresist of 1.1 ␮m thickness was used in negative mode along with a special lift-off photoresist (LOR2A) with 200 nm thickness underneath; with the T-shape observed as shown in Fig. 7. The resultant process gives a two layer stack and a “T-shaped” photoresist profile ideal for liftoff. The coating of aluminum in this layer was carried out with

Fig. 10. Probe station test setup used for measurements. Temperature readings taken from the sample stage are also given.

S.U. Senveli, O. Tigli / Sensors and Actuators B 196 (2014) 272–281

evaporation with a measured thickness of 125 nm. Final step in fabrication was lift-off using remover-pg solvent to remove the photoresist/metal using a 20 kHz ultrasonic buzzer for 5 min followed by another 5 min of replenished solution and cleaning after the samples were left to sit still in solution for 1 h. Overview of processed samples and fully formed IDT features are shown in Fig. 8. In order to be able to take measurements with liquids in microcavities, the issues of dispensing liquids in microcavities in such small volumes had to be resolved. It was experimentally verified that DI water was extremely volatile when produced in small amounts which renders it unusable. Surface tension of oils proved it to be difficult for the liquids to stay together during placement into cavities. Some other types of chemicals such as photoresists tend to harden and change to solid phase very quickly when in small amounts. On the other hand, glycerin based mixtures, due to the rather viscous nature, were easier to manipulate. For this reason, we used varying concentrations of mixtures of glycerin and DI water to demonstrate the operation of our microcavity based SAW sensors. For mixing purposes, glycerin and DI water were placed in containers after measuring with micropipettes, and then mixed using an automatic vortexer system for 5 min. This process was followed by ultrasonication of the containers for 10 min. The obtained mixtures contained no heterogeneities or air bubbles observable by a standard optical microscope. The concentrations of glycerin/water mixtures that were successfully dispensed in the microcavities are given in Table 1 along with their calculated densities and sound velocities. The application of liquids proved challenging when traditional methods are used such as syringe dispensing with smallest dispensable droplet sizes larger than 500 ␮m. For this reason, a standard DC probe measurement probe with 5 ␮m tip size was used to dispense the liquid into the microcavity with the help of 3-axis micropositioners as shown in Fig. 9. Due to surface tension of liquids on the probe tips and reliable dispensation in cavities, only 60–90% glycerin could be captured and operated with the probe tips. The test setup used for measurements is given in Fig. 10. To prevent false readings due to temperature effects, the temperature of the vacuum chuck of the probe station was continuously monitored during the testing phase. The maximum change of temperature is 0.3 ◦ C and using measured values of SAW delay coefficient for Y-Z lithium niobate (92 ppm/◦ C) [17], this translates only to approximately 5.8 kHz of frequency shift in our measurements which can be considered as a source of error in the collected frequency data. On the other hand, ST cut quartz has a SAW delay coefficient of zero around room temperature, meaning temperature does not incur measurable errors for ST-X quartz SAW devices in our measurements. The only source of error for frequency measurements of quartz samples is the uncertainty in the measurement due to the frequency conversion of the network analyzer.

4. Results and discussion The shape of the liquids when placed inside the microcavities was an important parameter which needed to be validated by surface angle measurements. It was seen that 90% glycerine has a surface contact angle of ∼40◦ with a hemispherical profile when dispensed on a regular lithium niobate surface. However, when placed into the microcavity, the liquid that peeks out does not form the shape of a hemisphere although it has a contact angle of ∼10◦ for 1.2 ␮m deep microcavities. This can be considered as a small deviation from the simulated model. Similarly, the same liquid on regular quartz has a contact angle of ∼40◦ but when placed in microcavities it seeps into the 8.5 ␮m deep microcavities and there is no hemispherical part on top. The results of the surface studies are shown in Fig. 11. This justifies that the use of a flat top instead of a spherical

279

Fig. 11. Microscope images of a separate set of microcavities used for optimizing the etching and the droplet dispensation process. (a) Top view of quartz substrate with 8.5 ␮m deep microcavities, (b) side view of the same area showing filled and empty microcavities and comparison to a droplet on the surface, (c) side view of lithium niobate with 1.2 ␮m deep cavities.

cap in the simulated models is an acceptable approximation. It is also seen by examining the dispensed droplets at different angles with a camera that the microcavities accept just enough volumes of liquid to fill them thanks to the surface tension acting on them. The amounts of liquids in the microcavities are estimated as 1 pL and 4 pL in lithium niobate and quartz, respectively. SAW delay lines with a wavelength of 16 ␮m and varying aperture sizes at passband peak frequencies of 214 MHz on Y-Z cut lithium niobate and 197 MHz ST-X cut quartz were used for liquid interrogation. The error with respect to simulations on the peak frequencies is 0.7% and 2.0%, respectively. No windowing techniques were used; all structures have constant overlap and a metallization ratio of unity. The microcavities with diameters of 24 ␮m are etched in the middle of the delay line, at equal distances from phase centers of the input and output IDTs. Sample testing was carried out one device at a time. Mainly, acetone and IPA were applied to clean the residues from the previous liquid. It was visually confirmed using the optical microscope that the microcavities were clean and empty of the previous liquid before dispensing the next liquid. The evaporation of liquid samples in the microcavities is considered to be negligible due to the low vapor pressures of glycerin and its mixtures with water. Nonetheless, microcavities were observed before and after measurements and compared to make sure they were filled. Insertion loss data was obtained using an Agilent E5061-B network analyzer and RF probes in only the frequency domain. The input power applied to the devices was −10 dBm. Intermediate frequency bandwidth (IFBW) was selected to a low value of 300 Hz to ensure little or no leakage from the local oscillator into the output. Choosing a low IFBW also causes the measurements to take considerably longer but maintains consistent data. Furthermore, data was averaged eight times during

280

S.U. Senveli, O. Tigli / Sensors and Actuators B 196 (2014) 272–281

Fig. 13. Compilation of measurement results (a) lithium niobate and (b) quartz.

Fig. 12. Insertion loss measurements from (a) lithium niobate and (b) quartz, with empty and filled microcavities. Liquids 1–4 are 60%, 70%, 80%, and 90% glycerin, respectively.

collection to make sure the results were consistent, noise level was limited, and any transient effects were not effective. The insertion loss measurements of a lithium niobate sample with 1.2 ␮m deep microcavities and a quartz substrate with 8.5 ␮m deep microcavities are shown in Fig. 12. Analysis of the data was carried out using Matlab to find the peak frequencies and the magnitude of insertion loss, and the phase shift data was averaged for accuracy. The compilations of data from both types of devices are given in Fig. 13. The error bars represent the standard deviation from five measurements and the temperature change where applicable (for the quartz substrate, the effect of temperature has been assumed to be negligible). The results show that the relation with phase difference is maintained in both conditions. On the other hand, frequency shifts do not exhibit a high correlation for deep microcavity conditions. Judging by the results from both types of devices and as expected from simulations, insertion loss magnitude is not a reliable parameter for detection as it varies to a great extent with the exact shape of the microcavity, and in turn the portion of SAW converted into bulk

waves. The effect of mass loading is more evident with the lithium niobate sample as the microcavity is shallower which approximates closer to a more standard delay line condition. The frequency drop increases with increased loading as expected. The slope after lin√ ear fitting is found as −16 kHz/(kg/m2 s) or −7.7 kHz/(% glycerin in the mixture). The phase difference is also larger than predicted with the simulations as the profile of the droplet differs by a small amount over the open top part of the microcavity. The fit for phase √ difference slope gives −1.3◦ /(kg/m2 s) or −0.63◦ /(% glycerin in the mixture). On the other hand, frequency drop readings exhibit much smaller differences and are not as accurate for the deeper microcavity on quartz. This is likely due to loading effects not effectively accounted for by the simulations. Correlation for frequency shifts is not adequate. However, the phase difference shows the same type of dependency. As the acoustic impedance of the liquid increases from L1 to L4, it starts getting closer to or matching that of the substrate. A negative slope relation exists between the liquid acoustic impedance and phase difference for the analyzed geometries as found in the FEM analysis. The fit curve for the quartz phase change √ gives −0.26◦ /(kg/m2 s) or −0.13◦ /(% glycerin in the mixture). As mentioned before, rather than a direct measurement; this is a result of the resonant microcavity condition which occurs due to the

S.U. Senveli, O. Tigli / Sensors and Actuators B 196 (2014) 272–281

specific microcavity geometry on the substrate. The shifts in phase are an indirect result of the liquid acoustic impedance. The differences between FEM studies and measurement results in both cases are attributed to the difference between constants used in simulations and the real samples. Also, the constants that we are using for our mixtures are based on those taken from liquids at much lower frequencies (usually in the kHz range) since higher frequency measurements were not available. Frequencies that we are using in our experiments are in the order of 200 MHz for very small sample sizes, unlike the other studies on the topic; therefore the constants are likely to deviate from tabulated values. Furthermore, the simulations assume exact dimensions that are multiples of the wavelength, deviations from these values also cause uncertainties especially in lithium niobate with shallow microcavities. The smallest density viscosity products differentiated with the sys√ tems are 1.9 kg/m2 s as observed from the phase difference. The lithium niobate case also gives a large frequency shift data which shows that shallow microcavities are useful for trapping liquids for measurement with less of the surface wave lost to bulk waves. On the other hand, the measurements with quartz demonstrate that deeper microcavities can be used for full interaction with the surface waves although losses are more significant (partly due to losses to the bulk); therefore the signal levels are lower. Errors are also larger with quartz due to higher magnitudes of insertion loss.

5. Conclusion Microcavities etched in delay lines are proposed as a method of interrogating minute amounts of liquids. It was seen by FEM simulations that the computationally complicated system can be modeled through coupling a piezoelectric substrate and the acoustic properties of a liquid. For testing purposes, SAW sensors were prepared on Y-Z lithium niobate and ST-X quartz substrates and etched to form microcavities. Liquid mixtures with varying glycerin/DI water contents were prepared for testing and then dispensed into the microcavities using probes and micropositioners. The characteristics expected from FEM simulations were also observed from measurements, and the proposed system with a microcavity has been differentiated from regular mass loading mode operation. The behavior of the system becomes closer to that of a mass loaded delay line if the microcavities are shallow and the effect of microcavities is more clearly observed when microcavities are etched deeper as relation is maintained between the liquid and the phase. Lithium niobate proves a higher sensitivity due to its higher coupling constant but can only support small depths of microcavities. For more controlled volume testing, quartz with specific microcavity depths can be considered. Materials with higher acoustic impedances placed in microcavities cause smaller shifts in the phase of the output insertion loss for the considered geometries. The phase shift is expected to operate in a cyclic fashion after some point in the density–viscosity product, and sets an upper limit to the dynamic range of the system. As a conclusion, liquids with volumes less than 5 pL have been analyzed using the setup, and the minimum detectable density-viscosity product difference between √ measurements was 1.9 kg/m2 s. It is seen that the presented microcavity approach can serve as a valuable tool that can be further improved for interrogating and differentiating between minute amounts of materials. Rayleigh waves are used in this study due to their well-established capability for coupling to liquids. While the first approach to this method proved successful, further research on this topic is necessary to make use of other types of surface waves to compare their performances and to improve the transducer designs in order to obtain the most efficient one for the microcavity approach outlined in this paper. It is proposed by the authors that this method can be further extended

281

to different materials, like solids or quasi-solid specimen such as biological cells with specialized design of IDTs. Acknowledgment Support from the National Science Foundation (NSF) under Grant No. ECCS-1349245 is gratefully acknowledged by the authors. References [1] M. Hoummady, A. Campitelli, W. Wlodarski, Acoustic wave sensors: design, sensing mechanisms and applications, Smart Materials and Structures 6 (1997) 647. [2] K. Lange, B.E. Rapp, M. Rapp, Surface acoustic wave biosensors: a review, Analytical and Bioanalytical Chemistry 391 (2008) 1509–1519. [3] O. Tigli, L. Bivona, P. Berg, M.E. Zaghloul, Fabrication and characterization of a surface-acoustic-wave biosensor in CMOS technology for cancer biomarker detection, IEEE Transactions on Biomedical Circuits and Systems 4 (2010) 62–73. [4] C. Wu, L. Du, D. Wang, L. Wang, L. Zhao, P. Wang, A novel surface acoustic wavebased biosensor for highly sensitive functional assays of olfactory receptors, Biochemical and Biophysical Research Communications 407 (2011) 18–22. [5] A. Sadek, W. Wlodarski, K. Shin, R. Kaner, K. Kalantar-Zadeh, A polyaniline/WO3 nanofiber composite-based ZnO/64(YX LiNbO3 SAW hydrogen gas sensor, Synthetic Metals 158 (2008) 29–32. [6] I. Gammoudi, H. Tarbague, A. Othmane, D. Moynet, D. Rebière, R. Kalfat, et al., Love-wave bacteria-based sensor for the detection of heavy metal toxicity in liquid medium, Biosensors and Bioelectronics 26 (2010) 1723–1726. [7] G. Lindner, Sensors and actuators based on surface acoustic waves propagating along solid–liquid interfaces, Journal of Physics D: Applied Physics 41 (2008) 123002. [8] V. Raimbault, D. Rebière, C. Dejous, A microfluidic surface acoustic wave sensor platform: application to high viscosity measurements, Materials Science and Engineering C 28 (2008) 759–764. [9] R. Lee, J. Vetelino, P. Clarke, A. Roy, J. Turner, Prototype microwave acoustic fluid sensors, in: Ultrasonics Symposium, 1988 Proceedings, IEEE, 1988, pp. 543–548. [10] Z. Wang, J. Zhe, Recent advances in particle and droplet manipulation for lab-on-a-chip devices based on surface acoustic waves, Lab Chip 11 (2011) 1280–1285. [11] Z. Rácz, M. Cole, J. Gardner, S. Pathak, M. Jordan, R. Challiss, Cell-based surface acoustic wave resonant microsensor for biomolecular agent detection, in: Solid-State Sensors, Actuators and Microsystems Conference (TRANSDUCERS), IEEE 16th International, 2011, pp. 2168–2171. [12] S. Cular, S.K.R.S. Sankaranarayanan, V.R. Bhethanabotla, Enhancing effects of microcavities on shear-horizontal surface acoustic wave sensors: a finite element simulation study, Applied Physics Letters 92 (2008) 244104–244113. [13] A. Wixforth, C. Strobl, C. Gauer, A. Toegl, J. Scriba, Z.V. Guttenberg, Acoustic manipulation of small droplets, Analytical and Bioanalytical Chemistry 379 (2004) 982–991. [14] H. Mechkour, The exact expressions for the roots of Rayleigh wave equation, in: Proceedings of the 2nd International Colloquium of Mathematics in Engineering and Numerical Physics, Bucharest, Romania, 2003. [15] N.S. Cheng, Formula for the viscosity of a glycerol–water mixture, Industrial and Engineering Chemistry Research 47 (2008) 3285–3288. [16] L. Elvira-Segura, Ultrasonic velocity measurement in liquids at low frequencies (<50 kHz) for milliliter samples, IEEE Transactions on Ultrasonics, Ferroelectrics and Frequency Control 48 (2001) 632–637. [17] R. Ward, Temperature coefficients of SAW delay and velocity for Y-cut and rotated LiNbO3 , IEEE Transactions on Ultrasonics, Ferroelectrics and Frequency Control 37 (1990) 481–483.

Biographies Sukru U. Senveli is currently pursuing his Ph.D. degree at University of Miami, Department of Electrical and Computer Engineering. He received the B.S. and M.S. degrees in Physics and Electrical and Electronics Engineering, from Middle East Technical University, Turkey. He conducted research on MEMS and integrated microdevices at METU MEMS Research & Application Center, for over four years. Currently, his main research interests include design and implementation of novel biomedical micro/nanosensors and their readouts, MEMS/NEMS design and fabrication, characterization methods, microfluidics, and electro-optics. Onur Tigli is an Assistant Professor of Electrical and Computer Engineering, and Dr. John T. McDonald Foundation Biomedical Nanotechnology Institute of University of Miami (BioNIUM). He also holds a secondary appointment at the Department of Pathology, Miller School of Medicine. Dr. Tigli received the B.S. degree in Electrical and Electronics Engineering from Middle East Technical University, Turkey, and the M.S. and D.Sc. degrees in Computer Engineering from George Washington University, Washington D.C. His primary research interest is in the field of bioMEMS/NEMS and biomedical nanotechnology to develop smart point-of-care diagnostic tools for clinical applications. He also conducts research in the field of MEMS scale energy harvesters and their biomedical applications.